In2Photos
Platinum Member
- Mar 21, 2007
- 2,159
- 2,234
- 136
OK, let me reword what I meant.Your post does not make sense. The 9700x has a max boost clock speed of 5500mhz. The 9800x3D has a boost in the video I posted of 5225mhz. The performance improvement of the 3D v-cache chips comes from the additional L3 cache stacked on the chips.
From Zen 4 to Zen 5 we all agree that the non-X3D parts did not have much of a performance uplift except in some specific use cases, correct? And this was despite the fact that there was a node improvement, correct? The 9700X and 9600X had a slight max frequency boost. As a refresher here are the Zen4 and Zen 5 non-X3D specs:
7950X, 5.7GHz max, 170W TDP >>>>> 9950X, 5.7GHz, 170W TDP
7900X, 5.6GHz max, 170W TDP >>>>> 9900X, 5.6GHz, 120W TDP
7700X, 5.4GHz max, 105W TDP >>>>> 9700X, 5.5GHz, 65/105W TDP
7600X, 5.3GHz max, 105W TDP >>>>> 9600X, 5.4GHz max, 65W TDP
The Zen5 skus got a little more efficient as we saw in reviews, but not much in performance, again, except for some very specific workloads.
Performance: N4P offers an 11% performance boost over N5
Power efficiency: N4P is 22% more power efficient than N5
Transistor density: N4P has a 6% higher transistor density than N5
6% higher performance vs N4
I think the issue here is you keep thinking that all that happened between Zen4 and Zen5 was better silicon, but that isn't the case. There is a lot more behind the scenes that changed. That's why we can't just apply that 11% and 22% improvement. It's like saying a car should see a 10% reduction in 0-60 times from year to year because the engine got 10% more horsepower. But you have to look at the car as a whole. There are many components that have to put that power to the ground. There could be styling changes, weight increases, gear changes, etc inhibiting that extra 10% from showing up.
That's why I am trying to figure out why the 9800x3D uses 10-20% more power while gaming than the 7800x3D which uses N5 silicon. That is real world observed power use while gaming and not paper numbers. I included the 6% higher performance number of N4P vs. N4 as well. N4P is very good silicon and the power use does not make sense. The X3D chips were voltage restricted in the past vs. standard X chips. So the power usage should be equal to or less. With N4P silicon efficiency gains should put the power usage at 7800x3D numbers or less. The 9800x3D clocks are less than 5% higher than the 7800x3D. Well within the TSMC statistical performance gains (N4P) through silicon alone.
Maybe AMD will release bios updates addressing the higher power use in the 9800x3D.
If we apply the same performance/efficiency increase we saw on the non-X3D chips from Zen 4 to Zen5 to the 7800X3D, the 9800X3D would likely only have seen a 3%-5% increase in performance as well, possibly with a slight power efficiency increase. But the 9800X3D saw the biggest change in frequency, >200Mhz. And the price to pay for that is power. It isn't free. What happened to the 9700X when the TDP was increased from 65W to 105W? We saw a rise in synthetic benchmarks, not much in gaming. The chip was already operating in the optimal performance range of the v/f curve. Once you step outside of that the increase of performance per watt is so small.
Now, what would be interesting to see is if you back the 9800X3D down from 120W to say 100W or 95W or limit the max frequency to 5.1GHz how much performance does it lose?
To me, Zen 5 is still rather lackluster in it's performance over Zen4 when you compare what AMD has done in previous generations. I may get flamed for saying that, but I think before the release of the 9800X3D everyone would have agreed (except for Mark and some others whose specific use case saw a rather large leap in performance). I think your point is something along these lines. You expected more from Zen5 simply because of the better silicon.